Face Generation¶

In this project, you'll use generative adversarial networks to generate new images of faces.

Get the Data¶

You'll be using two datasets in this project:

MNIST
CelebA

Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.

If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".

data_dir = './data'

# FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
#data_dir = '/input'


"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper

helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)

Downloading celeba: 0.00B [00:00, ?B/s]

Found mnist Data

Downloading celeba: 1.44GB [02:38, 9.08MB/s]

Extracting celeba...

Explore the Data¶

MNIST¶

As you're aware, the MNIST dataset contains images of handwritten digits. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot

mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')

<matplotlib.image.AxesImage at 0x7f88ea0224a8>

CelebA¶

The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))

<matplotlib.image.AxesImage at 0x7f88670c2cf8>

Preprocess the Data¶

Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.

The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).

Build the Neural Network¶

You'll build the components necessary to build a GANs by implementing the following functions below:

model_inputs
discriminator
generator
model_loss
model_opt
train

Check the Version of TensorFlow and Access to GPU¶

This will check to make sure you have the correct version of TensorFlow and access to a GPU

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf

# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))

TensorFlow Version: 1.0.0
Default GPU Device: /gpu:0

Input¶

Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:

Real input images placeholder with rank 4 using image_width, image_height, and image_channels.
Z input placeholder with rank 2 using z_dim.
Learning rate placeholder with rank 0.

Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)

import problem_unittests as tests

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data, learning rate)
    """
    # TODO: Implement Function
    
    input_real = tf.placeholder(tf.float32, (None, image_height, image_width, image_channels), name="input_real")
    input_z = tf.placeholder(tf.float32, (None, z_dim), name="input_z")
    learning_rate = tf.placeholder(tf.float32, (None), name="learning_rate")
    
    
    return input_real, input_z, learning_rate


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)

Tests Passed

Discriminator¶

Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variabes in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the generator, tensor logits of the generator).

def discriminator(images, reuse=False, alpha=0.01):
    """
    Create the discriminator network
    :param image: Tensor of input image(s)
    :param reuse: Boolean if the weights should be reused
    :return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
    """
    with tf.variable_scope('discriminator', reuse=reuse):
#         h1 = tf.reshape(images, (-1, 28*28*int(images.shape[-1])))
        
#         h1 = tf.layers.dense(h1, 1024, activation=None)
#         h1 = tf.maximum(alpha * h1, h1)
        
#         h2 = tf.layers.dense(h1, 512, activation=None)
#         h2 = tf.maximum(alpha * h2, h2)
        
#         logits = tf.layers.dense(h2, 1, activation=None)
#         out = tf.sigmoid(logits)
        
#         return out, logits

        size = 256
    
        x1 = tf.layers.conv2d(images, size/4, 5, strides=2, padding='same')
        x1 = tf.maximum(alpha * x1, x1)
        
        x2 = tf.layers.conv2d(x1, size/2, 5, strides=2, padding='same')
        x2 = tf.layers.batch_normalization(x2, training=True)
        x2 = tf.maximum(alpha * x2, x2)
        
        x3 = tf.layers.conv2d(x2, size, 5, strides=2, padding='same')
        x3 = tf.layers.batch_normalization(x3, training=True)
        x3 = tf.maximum(alpha * x3, x3)
    
        x3f = tf.reshape(x3, (-1, 4*4*size))
        
        logits = tf.layers.dense(x3f, 1)
        output = tf.sigmoid(logits)
        
        return output, logits


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)

Tests Passed

Generator¶

Implement generator to generate an image using z. This function should be able to reuse the variabes in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.

def generator(z, out_channel_dim, is_train=True, alpha=0.01):
    """
    Create the generator network
    :param z: Input z
    :param out_channel_dim: The number of channels in the output image
    :param is_train: Boolean if generator is being used for training
    :return: The tensor output of the generator
    """
    with tf.variable_scope('generator', reuse=(not is_train)):
#         h1 = tf.layers.dense(z, 256, activation=None)
#         h1 = tf.maximum(alpha * h1, h1)
        
#         h2 = tf.layers.dense(h1, 512, activation=None)
#         h2 = tf.maximum(alpha * h2, h2)
        
#         # Logits and tanh output
#         logits = tf.layers.dense(h2, 28*28*out_channel_dim, activation=None)
#         out = tf.tanh(logits)
        
#         return tf.reshape(out, (-1, 28, 28, out_channel_dim))

        size = 256

        x1 = tf.layers.dense(z, 7*7*size)
        
        x1 = tf.reshape(x1, (-1, 7, 7, size))
        x1 = tf.layers.batch_normalization(x1, training=is_train)
        x1 = tf.maximum(alpha * x1, x1)
        
        x2 = tf.layers.conv2d_transpose(x1, int(size/2), 5, strides=2, padding='same')
        x2 = tf.layers.batch_normalization(x2, training=is_train)
        x2 = tf.maximum(alpha * x2, x2)
        
        x3 = tf.layers.conv2d_transpose(x2, int(size/4), 5, strides=2, padding='same')
        x3 = tf.layers.batch_normalization(x3, training=is_train)
        x3 = tf.maximum(alpha * x3, x3)
        
        logits = tf.layers.conv2d_transpose(x3, out_channel_dim, 5, strides=1s, padding='same')
        output = tf.tanh(logits) * 0.5
        
        return output


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)

Tests Passed

Loss¶

Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:

discriminator(images, reuse=False)
generator(z, out_channel_dim, is_train=True)

def model_loss(input_real, input_z, out_channel_dim, alpha=0.01):
    """
    Get the loss for the discriminator and generator
    :param input_real: Images from the real dataset
    :param input_z: Z input
    :param out_channel_dim: The number of channels in the output image
    :return: A tuple of (discriminator loss, generator loss)
    """
    g_model = generator(input_z, out_channel_dim, is_train=True, alpha=alpha)
    
    d_model_real, d_logits_real = discriminator(input_real, reuse=False, alpha=alpha)
    d_model_fake, d_logits_fake = discriminator(g_model, reuse=True, alpha=alpha)

    d_loss_real = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_real, 
                                                labels=tf.ones_like(d_model_real) * (1 - 0.9)))
    d_loss_fake = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, 
                                                labels=tf.zeros_like(d_model_fake)))
    g_loss = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, 
                                                labels=tf.ones_like(d_model_fake)))

    d_loss = d_loss_real + d_loss_fake

    return d_loss, g_loss


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)

Tests Passed

Optimization¶

Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).

def model_opt(d_loss, g_loss, learning_rate, beta1):
    """
    Get optimization operations
    :param d_loss: Discriminator loss Tensor
    :param g_loss: Generator loss Tensor
    :param learning_rate: Learning Rate Placeholder
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :return: A tuple of (discriminator training operation, generator training operation)
    """
    t_vars = tf.trainable_variables()
    
    d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
    g_vars = [var for var in t_vars if var.name.startswith('generator')]

    with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
        d_train_opt = tf.train.AdamOptimizer(learning_rate, 
                                             beta1=beta1).minimize(d_loss, 
                                                                   var_list=d_vars)
        g_train_opt = tf.train.AdamOptimizer(learning_rate, 
                                             beta1=beta1).minimize(g_loss, 
                                                                   var_list=g_vars)

    return d_train_opt, g_train_opt


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)

Tests Passed

Neural Network Training¶

Show Output¶

Use this function to show the current output of the generator during training. It will help you determine how well the GANs is training.

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np

def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
    """
    Show example output for the generator
    :param sess: TensorFlow session
    :param n_images: Number of Images to display
    :param input_z: Input Z Tensor
    :param out_channel_dim: The number of channels in the output image
    :param image_mode: The mode to use for images ("RGB" or "L")
    """
    cmap = None if image_mode == 'RGB' else 'gray'
    z_dim = input_z.shape[-1]
    example_z = np.random.uniform(-1, 1, size=(n_images, z_dim))

    samples = sess.run(
        generator(input_z, out_channel_dim, False),
        feed_dict={input_z: example_z})

    images_grid = helper.images_square_grid(samples, image_mode)
    pyplot.imshow(images_grid, cmap=cmap)
    pyplot.show()

Train¶

Implement train to build and train the GANs. Use the following functions you implemented:

model_inputs(image_width, image_height, image_channels, z_dim)
model_loss(input_real, input_z, out_channel_dim)
model_opt(d_loss, g_loss, learning_rate, beta1)

Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.

def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode):
    """
    Train the GAN
    :param epoch_count: Number of epochs
    :param batch_size: Batch Size
    :param z_dim: Z dimension
    :param learning_rate: Learning Rate
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :param get_batches: Function to get batches
    :param data_shape: Shape of the data
    :param data_image_mode: The image mode to use for images ("RGB" or "L")
    """
    
    losses = []
    batch_i = 0
    
    batches_n = int(data_shape[0] / batch_size) * epoch_count
    
    input_real, input_z, learn_rate = model_inputs(data_shape[2], data_shape[1], data_shape[3], z_dim)
    d_loss, g_loss = model_loss(input_real, input_z, data_shape[3], alpha=0.01)
        
    d_opt, g_opt = model_opt(d_loss, g_loss, learning_rate, beta1)
    
    
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        for epoch_i in range(epoch_count):
            for batch_images in get_batches(batch_size):
                batch_i = batch_i + 1
                batch_z = np.random.uniform(-1, 1, size=(batch_size, z_dim))
                
                _ = sess.run(d_opt, feed_dict={input_real: batch_images, input_z: batch_z, learn_rate: learning_rate})
                _ = sess.run(g_opt, feed_dict={input_z: batch_z, input_real: batch_images, learn_rate: learning_rate})

                if batch_i % 1 == 0:
                    train_loss_d = d_loss.eval({input_z: batch_z, input_real: batch_images})
                    train_loss_g = g_loss.eval({input_z: batch_z})

                    print("Epoch {}/{} - ".format(epoch_i + 1, epochs),
                          "Batch {}/{} - ".format(batch_i, batches_n),
                          "Discriminator loss: {:.4f} - ".format(train_loss_d),
                          "Generator loss: {:.4f} - ".format(train_loss_g))
                    # Save losses to view after training
                    losses.append((train_loss_d, train_loss_g))

                if batch_i % 20 == 0:
                    show_generator_output(sess, 16, input_z, data_shape[3], data_image_mode)

MNIST¶

Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.

batch_size = 128
z_dim = 128
learning_rate = 0.0002
beta1 = 0.4


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

Epoch 1/2 -  Batch 1/936 -  Discriminator loss: 1.0075 -  Generator loss: 1.0218 - 
Epoch 1/2 -  Batch 2/936 -  Discriminator loss: 0.7437 -  Generator loss: 2.1129 - 
Epoch 1/2 -  Batch 3/936 -  Discriminator loss: 0.6531 -  Generator loss: 2.4898 - 
Epoch 1/2 -  Batch 4/936 -  Discriminator loss: 0.5533 -  Generator loss: 2.4544 - 
Epoch 1/2 -  Batch 5/936 -  Discriminator loss: 0.4594 -  Generator loss: 2.5052 - 
Epoch 1/2 -  Batch 6/936 -  Discriminator loss: 0.4214 -  Generator loss: 2.8483 - 
Epoch 1/2 -  Batch 7/936 -  Discriminator loss: 0.3889 -  Generator loss: 3.2375 - 
Epoch 1/2 -  Batch 8/936 -  Discriminator loss: 0.3902 -  Generator loss: 3.1998 - 
Epoch 1/2 -  Batch 9/936 -  Discriminator loss: 0.3907 -  Generator loss: 3.1861 - 
Epoch 1/2 -  Batch 10/936 -  Discriminator loss: 0.3710 -  Generator loss: 3.5652 - 
Epoch 1/2 -  Batch 11/936 -  Discriminator loss: 0.3606 -  Generator loss: 3.8523 - 
Epoch 1/2 -  Batch 12/936 -  Discriminator loss: 0.3645 -  Generator loss: 3.7731 - 
Epoch 1/2 -  Batch 13/936 -  Discriminator loss: 0.3661 -  Generator loss: 3.6052 - 
Epoch 1/2 -  Batch 14/936 -  Discriminator loss: 0.3587 -  Generator loss: 3.9719 - 
Epoch 1/2 -  Batch 15/936 -  Discriminator loss: 0.3528 -  Generator loss: 4.0753 - 
Epoch 1/2 -  Batch 16/936 -  Discriminator loss: 0.3531 -  Generator loss: 4.2390 - 
Epoch 1/2 -  Batch 17/936 -  Discriminator loss: 0.3905 -  Generator loss: 3.0602 - 
Epoch 1/2 -  Batch 18/936 -  Discriminator loss: 0.3712 -  Generator loss: 3.6028 - 
Epoch 1/2 -  Batch 19/936 -  Discriminator loss: 0.3626 -  Generator loss: 3.6695 - 
Epoch 1/2 -  Batch 20/936 -  Discriminator loss: 0.3673 -  Generator loss: 3.5983 -

Epoch 1/2 -  Batch 21/936 -  Discriminator loss: 0.3841 -  Generator loss: 3.1868 - 
Epoch 1/2 -  Batch 22/936 -  Discriminator loss: 0.4002 -  Generator loss: 3.0315 - 
Epoch 1/2 -  Batch 23/936 -  Discriminator loss: 0.4024 -  Generator loss: 2.9280 - 
Epoch 1/2 -  Batch 24/936 -  Discriminator loss: 0.3795 -  Generator loss: 3.4183 - 
Epoch 1/2 -  Batch 25/936 -  Discriminator loss: 0.3897 -  Generator loss: 3.1469 - 
Epoch 1/2 -  Batch 26/936 -  Discriminator loss: 0.3929 -  Generator loss: 3.2482 - 
Epoch 1/2 -  Batch 27/936 -  Discriminator loss: 0.4165 -  Generator loss: 2.7197 - 
Epoch 1/2 -  Batch 28/936 -  Discriminator loss: 0.3933 -  Generator loss: 3.7229 - 
Epoch 1/2 -  Batch 29/936 -  Discriminator loss: 0.4396 -  Generator loss: 2.5654 - 
Epoch 1/2 -  Batch 30/936 -  Discriminator loss: 0.4144 -  Generator loss: 4.5094 - 
Epoch 1/2 -  Batch 31/936 -  Discriminator loss: 0.4207 -  Generator loss: 2.7501 - 
Epoch 1/2 -  Batch 32/936 -  Discriminator loss: 0.4097 -  Generator loss: 4.1321 - 
Epoch 1/2 -  Batch 33/936 -  Discriminator loss: 0.4196 -  Generator loss: 2.7710 - 
Epoch 1/2 -  Batch 34/936 -  Discriminator loss: 0.4089 -  Generator loss: 4.1727 - 
Epoch 1/2 -  Batch 35/936 -  Discriminator loss: 0.4376 -  Generator loss: 2.5828 - 
Epoch 1/2 -  Batch 36/936 -  Discriminator loss: 0.4172 -  Generator loss: 4.2539 - 
Epoch 1/2 -  Batch 37/936 -  Discriminator loss: 0.4286 -  Generator loss: 2.6158 - 
Epoch 1/2 -  Batch 38/936 -  Discriminator loss: 0.4019 -  Generator loss: 4.3420 - 
Epoch 1/2 -  Batch 39/936 -  Discriminator loss: 0.4088 -  Generator loss: 2.8260 - 
Epoch 1/2 -  Batch 40/936 -  Discriminator loss: 0.4024 -  Generator loss: 4.5988 -

Epoch 1/2 -  Batch 41/936 -  Discriminator loss: 0.4157 -  Generator loss: 2.7238 - 
Epoch 1/2 -  Batch 42/936 -  Discriminator loss: 0.3943 -  Generator loss: 4.3047 - 
Epoch 1/2 -  Batch 43/936 -  Discriminator loss: 0.4148 -  Generator loss: 2.7464 - 
Epoch 1/2 -  Batch 44/936 -  Discriminator loss: 0.4345 -  Generator loss: 5.0531 - 
Epoch 1/2 -  Batch 45/936 -  Discriminator loss: 0.4034 -  Generator loss: 2.8524 - 
Epoch 1/2 -  Batch 46/936 -  Discriminator loss: 0.3862 -  Generator loss: 3.5644 - 
Epoch 1/2 -  Batch 47/936 -  Discriminator loss: 0.3879 -  Generator loss: 3.1393 - 
Epoch 1/2 -  Batch 48/936 -  Discriminator loss: 0.3830 -  Generator loss: 3.5370 - 
Epoch 1/2 -  Batch 49/936 -  Discriminator loss: 0.3849 -  Generator loss: 3.2536 - 
Epoch 1/2 -  Batch 50/936 -  Discriminator loss: 0.3792 -  Generator loss: 3.6464 - 
Epoch 1/2 -  Batch 51/936 -  Discriminator loss: 0.3898 -  Generator loss: 3.0747 - 
Epoch 1/2 -  Batch 52/936 -  Discriminator loss: 0.3937 -  Generator loss: 4.1970 - 
Epoch 1/2 -  Batch 53/936 -  Discriminator loss: 0.4234 -  Generator loss: 2.6957 - 
Epoch 1/2 -  Batch 54/936 -  Discriminator loss: 0.4404 -  Generator loss: 4.8511 - 
Epoch 1/2 -  Batch 55/936 -  Discriminator loss: 0.4014 -  Generator loss: 2.9526 - 
Epoch 1/2 -  Batch 56/936 -  Discriminator loss: 0.3877 -  Generator loss: 3.7029 - 
Epoch 1/2 -  Batch 57/936 -  Discriminator loss: 0.4108 -  Generator loss: 2.7321 - 
Epoch 1/2 -  Batch 58/936 -  Discriminator loss: 0.3985 -  Generator loss: 3.9481 - 
Epoch 1/2 -  Batch 59/936 -  Discriminator loss: 0.4248 -  Generator loss: 2.5943 - 
Epoch 1/2 -  Batch 60/936 -  Discriminator loss: 0.4416 -  Generator loss: 4.5127 -

Epoch 1/2 -  Batch 61/936 -  Discriminator loss: 0.4118 -  Generator loss: 2.7487 - 
Epoch 1/2 -  Batch 62/936 -  Discriminator loss: 0.3880 -  Generator loss: 3.3131 - 
Epoch 1/2 -  Batch 63/936 -  Discriminator loss: 0.3901 -  Generator loss: 3.2467 - 
Epoch 1/2 -  Batch 64/936 -  Discriminator loss: 0.4008 -  Generator loss: 2.8782 - 
Epoch 1/2 -  Batch 65/936 -  Discriminator loss: 0.3913 -  Generator loss: 3.4603 - 
Epoch 1/2 -  Batch 66/936 -  Discriminator loss: 0.4062 -  Generator loss: 2.7921 - 
Epoch 1/2 -  Batch 67/936 -  Discriminator loss: 0.3914 -  Generator loss: 3.3627 - 
Epoch 1/2 -  Batch 68/936 -  Discriminator loss: 0.4162 -  Generator loss: 2.7147 - 
Epoch 1/2 -  Batch 69/936 -  Discriminator loss: 0.4166 -  Generator loss: 4.0372 - 
Epoch 1/2 -  Batch 70/936 -  Discriminator loss: 0.4465 -  Generator loss: 2.4228 - 
Epoch 1/2 -  Batch 71/936 -  Discriminator loss: 0.4445 -  Generator loss: 4.3200 - 
Epoch 1/2 -  Batch 72/936 -  Discriminator loss: 0.4169 -  Generator loss: 2.6646 - 
Epoch 1/2 -  Batch 73/936 -  Discriminator loss: 0.3930 -  Generator loss: 3.3043 - 
Epoch 1/2 -  Batch 74/936 -  Discriminator loss: 0.4026 -  Generator loss: 2.8365 - 
Epoch 1/2 -  Batch 75/936 -  Discriminator loss: 0.3950 -  Generator loss: 3.2592 - 
Epoch 1/2 -  Batch 76/936 -  Discriminator loss: 0.3991 -  Generator loss: 2.8610 - 
Epoch 1/2 -  Batch 77/936 -  Discriminator loss: 0.3998 -  Generator loss: 3.4147 - 
Epoch 1/2 -  Batch 78/936 -  Discriminator loss: 0.4333 -  Generator loss: 2.4929 - 
Epoch 1/2 -  Batch 79/936 -  Discriminator loss: 0.4962 -  Generator loss: 5.0929 - 
Epoch 1/2 -  Batch 80/936 -  Discriminator loss: 0.3947 -  Generator loss: 3.1408 -

Epoch 1/2 -  Batch 81/936 -  Discriminator loss: 0.4933 -  Generator loss: 2.1131 - 
Epoch 1/2 -  Batch 82/936 -  Discriminator loss: 0.6135 -  Generator loss: 6.1431 - 
Epoch 1/2 -  Batch 83/936 -  Discriminator loss: 0.4677 -  Generator loss: 4.6295 - 
Epoch 1/2 -  Batch 84/936 -  Discriminator loss: 0.5495 -  Generator loss: 1.9467 - 
Epoch 1/2 -  Batch 85/936 -  Discriminator loss: 0.5234 -  Generator loss: 5.2349 - 
Epoch 1/2 -  Batch 86/936 -  Discriminator loss: 0.4221 -  Generator loss: 4.1984 - 
Epoch 1/2 -  Batch 87/936 -  Discriminator loss: 0.4764 -  Generator loss: 2.2799 - 
Epoch 1/2 -  Batch 88/936 -  Discriminator loss: 0.4588 -  Generator loss: 4.5481 - 
Epoch 1/2 -  Batch 89/936 -  Discriminator loss: 0.3974 -  Generator loss: 3.5870 - 
Epoch 1/2 -  Batch 90/936 -  Discriminator loss: 0.4605 -  Generator loss: 2.3206 - 
Epoch 1/2 -  Batch 91/936 -  Discriminator loss: 0.4402 -  Generator loss: 4.3971 - 
Epoch 1/2 -  Batch 92/936 -  Discriminator loss: 0.3979 -  Generator loss: 3.4415 - 
Epoch 1/2 -  Batch 93/936 -  Discriminator loss: 0.4433 -  Generator loss: 2.4211 - 
Epoch 1/2 -  Batch 94/936 -  Discriminator loss: 0.4176 -  Generator loss: 4.1462 - 
Epoch 1/2 -  Batch 95/936 -  Discriminator loss: 0.3945 -  Generator loss: 3.3092 - 
Epoch 1/2 -  Batch 96/936 -  Discriminator loss: 0.4226 -  Generator loss: 2.5468 - 
Epoch 1/2 -  Batch 97/936 -  Discriminator loss: 0.4016 -  Generator loss: 3.7291 - 
Epoch 1/2 -  Batch 98/936 -  Discriminator loss: 0.3977 -  Generator loss: 3.0385 - 
Epoch 1/2 -  Batch 99/936 -  Discriminator loss: 0.3951 -  Generator loss: 2.8742 - 
Epoch 1/2 -  Batch 100/936 -  Discriminator loss: 0.3900 -  Generator loss: 3.2192 -

Epoch 1/2 -  Batch 101/936 -  Discriminator loss: 0.3927 -  Generator loss: 3.0041 - 
Epoch 1/2 -  Batch 102/936 -  Discriminator loss: 0.3924 -  Generator loss: 3.0033 - 
Epoch 1/2 -  Batch 103/936 -  Discriminator loss: 0.3889 -  Generator loss: 3.1645 - 
Epoch 1/2 -  Batch 104/936 -  Discriminator loss: 0.3899 -  Generator loss: 3.0718 - 
Epoch 1/2 -  Batch 105/936 -  Discriminator loss: 0.3910 -  Generator loss: 3.0707 - 
Epoch 1/2 -  Batch 106/936 -  Discriminator loss: 0.3888 -  Generator loss: 3.1507 - 
Epoch 1/2 -  Batch 107/936 -  Discriminator loss: 0.3871 -  Generator loss: 3.1109 - 
Epoch 1/2 -  Batch 108/936 -  Discriminator loss: 0.3846 -  Generator loss: 3.1941 - 
Epoch 1/2 -  Batch 109/936 -  Discriminator loss: 0.3885 -  Generator loss: 3.0567 - 
Epoch 1/2 -  Batch 110/936 -  Discriminator loss: 0.3903 -  Generator loss: 3.1211 - 
Epoch 1/2 -  Batch 111/936 -  Discriminator loss: 0.3855 -  Generator loss: 3.2764 - 
Epoch 1/2 -  Batch 112/936 -  Discriminator loss: 0.3962 -  Generator loss: 2.8679 - 
Epoch 1/2 -  Batch 113/936 -  Discriminator loss: 0.3838 -  Generator loss: 3.4197 - 
Epoch 1/2 -  Batch 114/936 -  Discriminator loss: 0.3940 -  Generator loss: 2.9317 - 
Epoch 1/2 -  Batch 115/936 -  Discriminator loss: 0.3891 -  Generator loss: 3.2589 - 
Epoch 1/2 -  Batch 116/936 -  Discriminator loss: 0.3916 -  Generator loss: 2.9932 - 
Epoch 1/2 -  Batch 117/936 -  Discriminator loss: 0.3937 -  Generator loss: 3.0242 - 
Epoch 1/2 -  Batch 118/936 -  Discriminator loss: 0.3878 -  Generator loss: 3.1505 - 
Epoch 1/2 -  Batch 119/936 -  Discriminator loss: 0.3851 -  Generator loss: 3.2350 - 
Epoch 1/2 -  Batch 120/936 -  Discriminator loss: 0.3997 -  Generator loss: 2.8181 -

Epoch 1/2 -  Batch 121/936 -  Discriminator loss: 0.3918 -  Generator loss: 3.5723 - 
Epoch 1/2 -  Batch 122/936 -  Discriminator loss: 0.4045 -  Generator loss: 2.7385 - 
Epoch 1/2 -  Batch 123/936 -  Discriminator loss: 0.3974 -  Generator loss: 3.5650 - 
Epoch 1/2 -  Batch 124/936 -  Discriminator loss: 0.4029 -  Generator loss: 2.7533 - 
Epoch 1/2 -  Batch 125/936 -  Discriminator loss: 0.3914 -  Generator loss: 3.5904 - 
Epoch 1/2 -  Batch 126/936 -  Discriminator loss: 0.4085 -  Generator loss: 2.6673 - 
Epoch 1/2 -  Batch 127/936 -  Discriminator loss: 0.3987 -  Generator loss: 3.7684 - 
Epoch 1/2 -  Batch 128/936 -  Discriminator loss: 0.4042 -  Generator loss: 2.7239 - 
Epoch 1/2 -  Batch 129/936 -  Discriminator loss: 0.4003 -  Generator loss: 3.6994 - 
Epoch 1/2 -  Batch 130/936 -  Discriminator loss: 0.3999 -  Generator loss: 2.8046 - 
Epoch 1/2 -  Batch 131/936 -  Discriminator loss: 0.3843 -  Generator loss: 3.3705 - 
Epoch 1/2 -  Batch 132/936 -  Discriminator loss: 0.3850 -  Generator loss: 3.1219 - 
Epoch 1/2 -  Batch 133/936 -  Discriminator loss: 0.3914 -  Generator loss: 2.9906 - 
Epoch 1/2 -  Batch 134/936 -  Discriminator loss: 0.3869 -  Generator loss: 3.1419 - 
Epoch 1/2 -  Batch 135/936 -  Discriminator loss: 0.3839 -  Generator loss: 3.2993 - 
Epoch 1/2 -  Batch 136/936 -  Discriminator loss: 0.3870 -  Generator loss: 3.0180 - 
Epoch 1/2 -  Batch 137/936 -  Discriminator loss: 0.3808 -  Generator loss: 3.3311 - 
Epoch 1/2 -  Batch 138/936 -  Discriminator loss: 0.3846 -  Generator loss: 3.0913 - 
Epoch 1/2 -  Batch 139/936 -  Discriminator loss: 0.3820 -  Generator loss: 3.1613 - 
Epoch 1/2 -  Batch 140/936 -  Discriminator loss: 0.3835 -  Generator loss: 3.2599 -

Epoch 1/2 -  Batch 141/936 -  Discriminator loss: 0.3875 -  Generator loss: 3.0011 - 
Epoch 1/2 -  Batch 142/936 -  Discriminator loss: 0.3831 -  Generator loss: 3.3148 - 
Epoch 1/2 -  Batch 143/936 -  Discriminator loss: 0.3907 -  Generator loss: 2.9101 - 
Epoch 1/2 -  Batch 144/936 -  Discriminator loss: 0.3872 -  Generator loss: 3.4493 - 
Epoch 1/2 -  Batch 145/936 -  Discriminator loss: 0.3926 -  Generator loss: 2.8881 - 
Epoch 1/2 -  Batch 146/936 -  Discriminator loss: 0.3977 -  Generator loss: 4.0699 - 
Epoch 1/2 -  Batch 147/936 -  Discriminator loss: 0.4189 -  Generator loss: 2.5670 - 
Epoch 1/2 -  Batch 148/936 -  Discriminator loss: 0.4235 -  Generator loss: 4.3304 - 
Epoch 1/2 -  Batch 149/936 -  Discriminator loss: 0.3932 -  Generator loss: 2.8970 - 
Epoch 1/2 -  Batch 150/936 -  Discriminator loss: 0.3862 -  Generator loss: 3.4037 - 
Epoch 1/2 -  Batch 151/936 -  Discriminator loss: 0.3944 -  Generator loss: 2.8555 - 
Epoch 1/2 -  Batch 152/936 -  Discriminator loss: 0.3923 -  Generator loss: 3.8297 - 
Epoch 1/2 -  Batch 153/936 -  Discriminator loss: 0.4018 -  Generator loss: 2.7542 - 
Epoch 1/2 -  Batch 154/936 -  Discriminator loss: 0.3955 -  Generator loss: 3.7133 - 
Epoch 1/2 -  Batch 155/936 -  Discriminator loss: 0.3988 -  Generator loss: 2.8192 - 
Epoch 1/2 -  Batch 156/936 -  Discriminator loss: 0.3886 -  Generator loss: 3.7717 - 
Epoch 1/2 -  Batch 157/936 -  Discriminator loss: 0.3985 -  Generator loss: 2.7973 - 
Epoch 1/2 -  Batch 158/936 -  Discriminator loss: 0.3932 -  Generator loss: 3.8990 - 
Epoch 1/2 -  Batch 159/936 -  Discriminator loss: 0.4042 -  Generator loss: 2.7408 - 
Epoch 1/2 -  Batch 160/936 -  Discriminator loss: 0.3917 -  Generator loss: 3.7544 -

Epoch 1/2 -  Batch 161/936 -  Discriminator loss: 0.3897 -  Generator loss: 2.9219 - 
Epoch 1/2 -  Batch 162/936 -  Discriminator loss: 0.3845 -  Generator loss: 3.2779 - 
Epoch 1/2 -  Batch 163/936 -  Discriminator loss: 0.3785 -  Generator loss: 3.2288 - 
Epoch 1/2 -  Batch 164/936 -  Discriminator loss: 0.3792 -  Generator loss: 3.3104 - 
Epoch 1/2 -  Batch 165/936 -  Discriminator loss: 0.3866 -  Generator loss: 2.9908 - 
Epoch 1/2 -  Batch 166/936 -  Discriminator loss: 0.3884 -  Generator loss: 3.8445 - 
Epoch 1/2 -  Batch 167/936 -  Discriminator loss: 0.4033 -  Generator loss: 2.7718 - 
Epoch 1/2 -  Batch 168/936 -  Discriminator loss: 0.3902 -  Generator loss: 3.7480 - 
Epoch 1/2 -  Batch 169/936 -  Discriminator loss: 0.3874 -  Generator loss: 2.9798 - 
Epoch 1/2 -  Batch 170/936 -  Discriminator loss: 0.3833 -  Generator loss: 3.5971 - 
Epoch 1/2 -  Batch 171/936 -  Discriminator loss: 0.3938 -  Generator loss: 2.8804 - 
Epoch 1/2 -  Batch 172/936 -  Discriminator loss: 0.3839 -  Generator loss: 3.7711 - 
Epoch 1/2 -  Batch 173/936 -  Discriminator loss: 0.3876 -  Generator loss: 2.9652 - 
Epoch 1/2 -  Batch 174/936 -  Discriminator loss: 0.3778 -  Generator loss: 3.6592 - 
Epoch 1/2 -  Batch 175/936 -  Discriminator loss: 0.3881 -  Generator loss: 2.9851 - 
Epoch 1/2 -  Batch 176/936 -  Discriminator loss: 0.3822 -  Generator loss: 3.5842 - 
Epoch 1/2 -  Batch 177/936 -  Discriminator loss: 0.3881 -  Generator loss: 2.9489 - 
Epoch 1/2 -  Batch 178/936 -  Discriminator loss: 0.3934 -  Generator loss: 4.1815 - 
Epoch 1/2 -  Batch 179/936 -  Discriminator loss: 0.4084 -  Generator loss: 2.7374 - 
Epoch 1/2 -  Batch 180/936 -  Discriminator loss: 0.4112 -  Generator loss: 4.3094 -

Epoch 1/2 -  Batch 181/936 -  Discriminator loss: 0.3905 -  Generator loss: 2.9664 - 
Epoch 1/2 -  Batch 182/936 -  Discriminator loss: 0.3780 -  Generator loss: 3.4045 - 
Epoch 1/2 -  Batch 183/936 -  Discriminator loss: 0.3787 -  Generator loss: 3.1859 - 
Epoch 1/2 -  Batch 184/936 -  Discriminator loss: 0.3789 -  Generator loss: 3.1886 - 
Epoch 1/2 -  Batch 185/936 -  Discriminator loss: 0.3788 -  Generator loss: 3.3500 - 
Epoch 1/2 -  Batch 186/936 -  Discriminator loss: 0.3758 -  Generator loss: 3.3409 - 
Epoch 1/2 -  Batch 187/936 -  Discriminator loss: 0.3835 -  Generator loss: 3.0858 - 
Epoch 1/2 -  Batch 188/936 -  Discriminator loss: 0.3801 -  Generator loss: 3.8306 - 
Epoch 1/2 -  Batch 189/936 -  Discriminator loss: 0.4000 -  Generator loss: 2.8130 - 
Epoch 1/2 -  Batch 190/936 -  Discriminator loss: 0.4182 -  Generator loss: 4.4663 - 
Epoch 1/2 -  Batch 191/936 -  Discriminator loss: 0.3906 -  Generator loss: 2.9554 - 
Epoch 1/2 -  Batch 192/936 -  Discriminator loss: 0.3801 -  Generator loss: 3.4963 - 
Epoch 1/2 -  Batch 193/936 -  Discriminator loss: 0.3848 -  Generator loss: 3.0764 - 
Epoch 1/2 -  Batch 194/936 -  Discriminator loss: 0.3777 -  Generator loss: 3.5555 - 
Epoch 1/2 -  Batch 195/936 -  Discriminator loss: 0.3845 -  Generator loss: 3.0280 - 
Epoch 1/2 -  Batch 196/936 -  Discriminator loss: 0.3782 -  Generator loss: 3.6870 - 
Epoch 1/2 -  Batch 197/936 -  Discriminator loss: 0.3894 -  Generator loss: 2.9347 - 
Epoch 1/2 -  Batch 198/936 -  Discriminator loss: 0.3959 -  Generator loss: 4.2005 - 
Epoch 1/2 -  Batch 199/936 -  Discriminator loss: 0.4051 -  Generator loss: 2.7536 - 
Epoch 1/2 -  Batch 200/936 -  Discriminator loss: 0.3994 -  Generator loss: 4.1426 -

CelebA¶

Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.

batch_size = 128
z_dim = 128
learning_rate = 0.0001
beta1 = 0.4


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 1

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

Epoch 1/1 -  Batch 1/1582 -  Discriminator loss: 1.3107 -  Generator loss: 0.5529 - 
Epoch 1/1 -  Batch 2/1582 -  Discriminator loss: 0.8683 -  Generator loss: 1.0690 - 
Epoch 1/1 -  Batch 3/1582 -  Discriminator loss: 0.7684 -  Generator loss: 1.3486 - 
Epoch 1/1 -  Batch 4/1582 -  Discriminator loss: 0.7182 -  Generator loss: 1.5084 - 
Epoch 1/1 -  Batch 5/1582 -  Discriminator loss: 0.6522 -  Generator loss: 1.6655 - 
Epoch 1/1 -  Batch 6/1582 -  Discriminator loss: 0.5907 -  Generator loss: 1.8739 - 
Epoch 1/1 -  Batch 7/1582 -  Discriminator loss: 0.5444 -  Generator loss: 1.9661 - 
Epoch 1/1 -  Batch 8/1582 -  Discriminator loss: 0.5292 -  Generator loss: 2.0743 - 
Epoch 1/1 -  Batch 9/1582 -  Discriminator loss: 0.5205 -  Generator loss: 2.1081 - 
Epoch 1/1 -  Batch 10/1582 -  Discriminator loss: 0.4997 -  Generator loss: 2.1716 - 
Epoch 1/1 -  Batch 11/1582 -  Discriminator loss: 0.5077 -  Generator loss: 2.1756 - 
Epoch 1/1 -  Batch 12/1582 -  Discriminator loss: 0.4789 -  Generator loss: 2.3697 - 
Epoch 1/1 -  Batch 13/1582 -  Discriminator loss: 0.4702 -  Generator loss: 2.3678 - 
Epoch 1/1 -  Batch 14/1582 -  Discriminator loss: 0.4597 -  Generator loss: 2.4626 - 
Epoch 1/1 -  Batch 15/1582 -  Discriminator loss: 0.4690 -  Generator loss: 2.4294 - 
Epoch 1/1 -  Batch 16/1582 -  Discriminator loss: 0.4561 -  Generator loss: 2.4587 - 
Epoch 1/1 -  Batch 17/1582 -  Discriminator loss: 0.4686 -  Generator loss: 2.4431 - 
Epoch 1/1 -  Batch 18/1582 -  Discriminator loss: 0.4630 -  Generator loss: 2.4346 - 
Epoch 1/1 -  Batch 19/1582 -  Discriminator loss: 0.4566 -  Generator loss: 2.5603 - 
Epoch 1/1 -  Batch 20/1582 -  Discriminator loss: 0.4654 -  Generator loss: 2.5315 -

Epoch 1/1 -  Batch 21/1582 -  Discriminator loss: 0.4398 -  Generator loss: 2.6189 - 
Epoch 1/1 -  Batch 22/1582 -  Discriminator loss: 0.4351 -  Generator loss: 2.6345 - 
Epoch 1/1 -  Batch 23/1582 -  Discriminator loss: 0.4236 -  Generator loss: 2.7878 - 
Epoch 1/1 -  Batch 24/1582 -  Discriminator loss: 0.4278 -  Generator loss: 2.8036 - 
Epoch 1/1 -  Batch 25/1582 -  Discriminator loss: 0.4203 -  Generator loss: 2.8254 - 
Epoch 1/1 -  Batch 26/1582 -  Discriminator loss: 0.4151 -  Generator loss: 2.9110 - 
Epoch 1/1 -  Batch 27/1582 -  Discriminator loss: 0.4099 -  Generator loss: 3.0467 - 
Epoch 1/1 -  Batch 28/1582 -  Discriminator loss: 0.4086 -  Generator loss: 3.1461 - 
Epoch 1/1 -  Batch 29/1582 -  Discriminator loss: 0.3979 -  Generator loss: 3.0897 - 
Epoch 1/1 -  Batch 30/1582 -  Discriminator loss: 0.4048 -  Generator loss: 3.0687 - 
Epoch 1/1 -  Batch 31/1582 -  Discriminator loss: 0.4041 -  Generator loss: 3.0727 - 
Epoch 1/1 -  Batch 32/1582 -  Discriminator loss: 0.4049 -  Generator loss: 3.0838 - 
Epoch 1/1 -  Batch 33/1582 -  Discriminator loss: 0.3955 -  Generator loss: 3.1384 - 
Epoch 1/1 -  Batch 34/1582 -  Discriminator loss: 0.4036 -  Generator loss: 3.1035 - 
Epoch 1/1 -  Batch 35/1582 -  Discriminator loss: 0.3948 -  Generator loss: 3.1702 - 
Epoch 1/1 -  Batch 36/1582 -  Discriminator loss: 0.3933 -  Generator loss: 3.4458 - 
Epoch 1/1 -  Batch 37/1582 -  Discriminator loss: 0.3884 -  Generator loss: 3.3311 - 
Epoch 1/1 -  Batch 38/1582 -  Discriminator loss: 0.3895 -  Generator loss: 3.3134 - 
Epoch 1/1 -  Batch 39/1582 -  Discriminator loss: 0.3877 -  Generator loss: 3.5417 - 
Epoch 1/1 -  Batch 40/1582 -  Discriminator loss: 0.3866 -  Generator loss: 3.3969 -

Epoch 1/1 -  Batch 41/1582 -  Discriminator loss: 0.3822 -  Generator loss: 3.5435 - 
Epoch 1/1 -  Batch 42/1582 -  Discriminator loss: 0.3907 -  Generator loss: 3.3540 - 
Epoch 1/1 -  Batch 43/1582 -  Discriminator loss: 0.3891 -  Generator loss: 3.2664 - 
Epoch 1/1 -  Batch 44/1582 -  Discriminator loss: 0.3952 -  Generator loss: 3.2199 - 
Epoch 1/1 -  Batch 45/1582 -  Discriminator loss: 0.3930 -  Generator loss: 3.2450 - 
Epoch 1/1 -  Batch 46/1582 -  Discriminator loss: 0.3835 -  Generator loss: 3.4149 - 
Epoch 1/1 -  Batch 47/1582 -  Discriminator loss: 0.3849 -  Generator loss: 3.3379 - 
Epoch 1/1 -  Batch 48/1582 -  Discriminator loss: 0.3863 -  Generator loss: 3.3220 - 
Epoch 1/1 -  Batch 49/1582 -  Discriminator loss: 0.3798 -  Generator loss: 3.4867 - 
Epoch 1/1 -  Batch 50/1582 -  Discriminator loss: 0.3709 -  Generator loss: 3.6170 - 
Epoch 1/1 -  Batch 51/1582 -  Discriminator loss: 0.3734 -  Generator loss: 3.7032 - 
Epoch 1/1 -  Batch 52/1582 -  Discriminator loss: 0.3704 -  Generator loss: 3.7015 - 
Epoch 1/1 -  Batch 53/1582 -  Discriminator loss: 0.3753 -  Generator loss: 3.5682 - 
Epoch 1/1 -  Batch 54/1582 -  Discriminator loss: 0.3708 -  Generator loss: 3.7530 - 
Epoch 1/1 -  Batch 55/1582 -  Discriminator loss: 0.3679 -  Generator loss: 3.8163 - 
Epoch 1/1 -  Batch 56/1582 -  Discriminator loss: 0.3739 -  Generator loss: 3.6899 - 
Epoch 1/1 -  Batch 57/1582 -  Discriminator loss: 0.3673 -  Generator loss: 3.9510 - 
Epoch 1/1 -  Batch 58/1582 -  Discriminator loss: 0.3723 -  Generator loss: 3.7530 - 
Epoch 1/1 -  Batch 59/1582 -  Discriminator loss: 0.3643 -  Generator loss: 3.9276 - 
Epoch 1/1 -  Batch 60/1582 -  Discriminator loss: 0.3715 -  Generator loss: 3.7029 -

Epoch 1/1 -  Batch 61/1582 -  Discriminator loss: 0.3660 -  Generator loss: 3.7616 - 
Epoch 1/1 -  Batch 62/1582 -  Discriminator loss: 0.3668 -  Generator loss: 3.8228 - 
Epoch 1/1 -  Batch 63/1582 -  Discriminator loss: 0.3649 -  Generator loss: 3.8844 - 
Epoch 1/1 -  Batch 64/1582 -  Discriminator loss: 0.3626 -  Generator loss: 3.8605 - 
Epoch 1/1 -  Batch 65/1582 -  Discriminator loss: 0.3591 -  Generator loss: 4.0243 - 
Epoch 1/1 -  Batch 66/1582 -  Discriminator loss: 0.3647 -  Generator loss: 3.8714 - 
Epoch 1/1 -  Batch 67/1582 -  Discriminator loss: 0.3607 -  Generator loss: 3.8913 - 
Epoch 1/1 -  Batch 68/1582 -  Discriminator loss: 0.3574 -  Generator loss: 4.1018 - 
Epoch 1/1 -  Batch 69/1582 -  Discriminator loss: 0.3607 -  Generator loss: 4.1008 - 
Epoch 1/1 -  Batch 70/1582 -  Discriminator loss: 0.3608 -  Generator loss: 3.8847 - 
Epoch 1/1 -  Batch 71/1582 -  Discriminator loss: 0.3665 -  Generator loss: 3.7289 - 
Epoch 1/1 -  Batch 72/1582 -  Discriminator loss: 0.3615 -  Generator loss: 3.8622 - 
Epoch 1/1 -  Batch 73/1582 -  Discriminator loss: 0.3602 -  Generator loss: 3.9603 - 
Epoch 1/1 -  Batch 74/1582 -  Discriminator loss: 0.3630 -  Generator loss: 3.8573 - 
Epoch 1/1 -  Batch 75/1582 -  Discriminator loss: 0.3527 -  Generator loss: 4.1662 - 
Epoch 1/1 -  Batch 76/1582 -  Discriminator loss: 0.3541 -  Generator loss: 4.2512 - 
Epoch 1/1 -  Batch 77/1582 -  Discriminator loss: 0.3626 -  Generator loss: 3.9557 - 
Epoch 1/1 -  Batch 78/1582 -  Discriminator loss: 0.3578 -  Generator loss: 4.0057 - 
Epoch 1/1 -  Batch 79/1582 -  Discriminator loss: 0.3551 -  Generator loss: 4.0399 - 
Epoch 1/1 -  Batch 80/1582 -  Discriminator loss: 0.3597 -  Generator loss: 3.9401 -

Epoch 1/1 -  Batch 81/1582 -  Discriminator loss: 0.3571 -  Generator loss: 4.0352 - 
Epoch 1/1 -  Batch 82/1582 -  Discriminator loss: 0.3563 -  Generator loss: 4.1264 - 
Epoch 1/1 -  Batch 83/1582 -  Discriminator loss: 0.3624 -  Generator loss: 3.7805 - 
Epoch 1/1 -  Batch 84/1582 -  Discriminator loss: 0.3574 -  Generator loss: 4.0755 - 
Epoch 1/1 -  Batch 85/1582 -  Discriminator loss: 0.3631 -  Generator loss: 3.7876 - 
Epoch 1/1 -  Batch 86/1582 -  Discriminator loss: 0.3634 -  Generator loss: 3.7537 - 
Epoch 1/1 -  Batch 87/1582 -  Discriminator loss: 0.3633 -  Generator loss: 3.8775 - 
Epoch 1/1 -  Batch 88/1582 -  Discriminator loss: 0.3597 -  Generator loss: 3.9536 - 
Epoch 1/1 -  Batch 89/1582 -  Discriminator loss: 0.3561 -  Generator loss: 4.1365 - 
Epoch 1/1 -  Batch 90/1582 -  Discriminator loss: 0.3597 -  Generator loss: 3.8791 - 
Epoch 1/1 -  Batch 91/1582 -  Discriminator loss: 0.3544 -  Generator loss: 4.1000 - 
Epoch 1/1 -  Batch 92/1582 -  Discriminator loss: 0.3561 -  Generator loss: 3.9594 - 
Epoch 1/1 -  Batch 93/1582 -  Discriminator loss: 0.3528 -  Generator loss: 4.3277 - 
Epoch 1/1 -  Batch 94/1582 -  Discriminator loss: 0.3599 -  Generator loss: 3.8470 - 
Epoch 1/1 -  Batch 95/1582 -  Discriminator loss: 0.3585 -  Generator loss: 3.7792 - 
Epoch 1/1 -  Batch 96/1582 -  Discriminator loss: 0.3606 -  Generator loss: 3.9934 - 
Epoch 1/1 -  Batch 97/1582 -  Discriminator loss: 0.3598 -  Generator loss: 4.0775 - 
Epoch 1/1 -  Batch 98/1582 -  Discriminator loss: 0.3597 -  Generator loss: 3.8228 - 
Epoch 1/1 -  Batch 99/1582 -  Discriminator loss: 0.3634 -  Generator loss: 4.5224 - 
Epoch 1/1 -  Batch 100/1582 -  Discriminator loss: 0.3602 -  Generator loss: 3.9217 -

Epoch 1/1 -  Batch 101/1582 -  Discriminator loss: 0.3581 -  Generator loss: 4.0831 - 
Epoch 1/1 -  Batch 102/1582 -  Discriminator loss: 0.3629 -  Generator loss: 3.9465 - 
Epoch 1/1 -  Batch 103/1582 -  Discriminator loss: 0.3557 -  Generator loss: 4.1245 - 
Epoch 1/1 -  Batch 104/1582 -  Discriminator loss: 0.3581 -  Generator loss: 4.1474 - 
Epoch 1/1 -  Batch 105/1582 -  Discriminator loss: 0.3569 -  Generator loss: 4.1166 - 
Epoch 1/1 -  Batch 106/1582 -  Discriminator loss: 0.3627 -  Generator loss: 3.8400 - 
Epoch 1/1 -  Batch 107/1582 -  Discriminator loss: 0.3578 -  Generator loss: 4.3916 - 
Epoch 1/1 -  Batch 108/1582 -  Discriminator loss: 0.3613 -  Generator loss: 3.9775 - 
Epoch 1/1 -  Batch 109/1582 -  Discriminator loss: 0.3600 -  Generator loss: 4.0099 - 
Epoch 1/1 -  Batch 110/1582 -  Discriminator loss: 0.3608 -  Generator loss: 4.1918 - 
Epoch 1/1 -  Batch 111/1582 -  Discriminator loss: 0.3595 -  Generator loss: 4.0683 - 
Epoch 1/1 -  Batch 112/1582 -  Discriminator loss: 0.3573 -  Generator loss: 4.2024 - 
Epoch 1/1 -  Batch 113/1582 -  Discriminator loss: 0.3601 -  Generator loss: 3.9445 - 
Epoch 1/1 -  Batch 114/1582 -  Discriminator loss: 0.3583 -  Generator loss: 4.5255 - 
Epoch 1/1 -  Batch 115/1582 -  Discriminator loss: 0.3629 -  Generator loss: 3.7944 - 
Epoch 1/1 -  Batch 116/1582 -  Discriminator loss: 0.3550 -  Generator loss: 4.3641 - 
Epoch 1/1 -  Batch 117/1582 -  Discriminator loss: 0.3589 -  Generator loss: 3.9744 - 
Epoch 1/1 -  Batch 118/1582 -  Discriminator loss: 0.3541 -  Generator loss: 4.1425 - 
Epoch 1/1 -  Batch 119/1582 -  Discriminator loss: 0.3567 -  Generator loss: 4.0829 - 
Epoch 1/1 -  Batch 120/1582 -  Discriminator loss: 0.3556 -  Generator loss: 3.9733 -

Epoch 1/1 -  Batch 121/1582 -  Discriminator loss: 0.3601 -  Generator loss: 3.9388 - 
Epoch 1/1 -  Batch 122/1582 -  Discriminator loss: 0.3623 -  Generator loss: 3.9720 - 
Epoch 1/1 -  Batch 123/1582 -  Discriminator loss: 0.3640 -  Generator loss: 3.6598 - 
Epoch 1/1 -  Batch 124/1582 -  Discriminator loss: 0.3645 -  Generator loss: 4.1710 - 
Epoch 1/1 -  Batch 125/1582 -  Discriminator loss: 0.3801 -  Generator loss: 3.2891 - 
Epoch 1/1 -  Batch 126/1582 -  Discriminator loss: 0.3670 -  Generator loss: 3.6806 - 
Epoch 1/1 -  Batch 127/1582 -  Discriminator loss: 0.3637 -  Generator loss: 3.9021 - 
Epoch 1/1 -  Batch 128/1582 -  Discriminator loss: 0.3750 -  Generator loss: 3.3816 - 
Epoch 1/1 -  Batch 129/1582 -  Discriminator loss: 0.3721 -  Generator loss: 3.7736 - 
Epoch 1/1 -  Batch 130/1582 -  Discriminator loss: 0.3842 -  Generator loss: 3.2082 - 
Epoch 1/1 -  Batch 131/1582 -  Discriminator loss: 0.3865 -  Generator loss: 4.2944 - 
Epoch 1/1 -  Batch 132/1582 -  Discriminator loss: 0.3739 -  Generator loss: 3.5792 - 
Epoch 1/1 -  Batch 133/1582 -  Discriminator loss: 0.3663 -  Generator loss: 3.8900 - 
Epoch 1/1 -  Batch 134/1582 -  Discriminator loss: 0.3717 -  Generator loss: 3.5371 - 
Epoch 1/1 -  Batch 135/1582 -  Discriminator loss: 0.3679 -  Generator loss: 3.9369 - 
Epoch 1/1 -  Batch 136/1582 -  Discriminator loss: 0.3698 -  Generator loss: 3.6633 - 
Epoch 1/1 -  Batch 137/1582 -  Discriminator loss: 0.3718 -  Generator loss: 3.5267 - 
Epoch 1/1 -  Batch 138/1582 -  Discriminator loss: 0.3698 -  Generator loss: 3.7072 - 
Epoch 1/1 -  Batch 139/1582 -  Discriminator loss: 0.3742 -  Generator loss: 3.4796 - 
Epoch 1/1 -  Batch 140/1582 -  Discriminator loss: 0.3780 -  Generator loss: 4.1290 -

Epoch 1/1 -  Batch 141/1582 -  Discriminator loss: 0.3866 -  Generator loss: 3.1920 - 
Epoch 1/1 -  Batch 142/1582 -  Discriminator loss: 0.3876 -  Generator loss: 4.1846 - 
Epoch 1/1 -  Batch 143/1582 -  Discriminator loss: 0.3794 -  Generator loss: 3.4988 - 
Epoch 1/1 -  Batch 144/1582 -  Discriminator loss: 0.3792 -  Generator loss: 3.3665 - 
Epoch 1/1 -  Batch 145/1582 -  Discriminator loss: 0.3836 -  Generator loss: 4.1680 - 
Epoch 1/1 -  Batch 146/1582 -  Discriminator loss: 0.3745 -  Generator loss: 3.4821 - 
Epoch 1/1 -  Batch 147/1582 -  Discriminator loss: 0.3758 -  Generator loss: 3.4486 - 
Epoch 1/1 -  Batch 148/1582 -  Discriminator loss: 0.3756 -  Generator loss: 3.6879 - 
Epoch 1/1 -  Batch 149/1582 -  Discriminator loss: 0.3800 -  Generator loss: 3.4749 - 
Epoch 1/1 -  Batch 150/1582 -  Discriminator loss: 0.3856 -  Generator loss: 3.2538 - 
Epoch 1/1 -  Batch 151/1582 -  Discriminator loss: 0.3802 -  Generator loss: 3.6270 - 
Epoch 1/1 -  Batch 152/1582 -  Discriminator loss: 0.3791 -  Generator loss: 3.4583 - 
Epoch 1/1 -  Batch 153/1582 -  Discriminator loss: 0.3792 -  Generator loss: 3.4759 - 
Epoch 1/1 -  Batch 154/1582 -  Discriminator loss: 0.3769 -  Generator loss: 3.4091 - 
Epoch 1/1 -  Batch 155/1582 -  Discriminator loss: 0.3765 -  Generator loss: 3.5901 - 
Epoch 1/1 -  Batch 156/1582 -  Discriminator loss: 0.3753 -  Generator loss: 3.6263 - 
Epoch 1/1 -  Batch 157/1582 -  Discriminator loss: 0.3753 -  Generator loss: 3.5847 - 
Epoch 1/1 -  Batch 158/1582 -  Discriminator loss: 0.3696 -  Generator loss: 4.1804 - 
Epoch 1/1 -  Batch 159/1582 -  Discriminator loss: 0.3631 -  Generator loss: 3.6733 - 
Epoch 1/1 -  Batch 160/1582 -  Discriminator loss: 0.3646 -  Generator loss: 3.9109 -

Epoch 1/1 -  Batch 161/1582 -  Discriminator loss: 0.3679 -  Generator loss: 3.7685 - 
Epoch 1/1 -  Batch 162/1582 -  Discriminator loss: 0.3825 -  Generator loss: 3.2341 - 
Epoch 1/1 -  Batch 163/1582 -  Discriminator loss: 0.3810 -  Generator loss: 4.0346 - 
Epoch 1/1 -  Batch 164/1582 -  Discriminator loss: 0.3808 -  Generator loss: 3.2647 - 
Epoch 1/1 -  Batch 165/1582 -  Discriminator loss: 0.3741 -  Generator loss: 4.0493 - 
Epoch 1/1 -  Batch 166/1582 -  Discriminator loss: 0.3748 -  Generator loss: 3.3455 - 
Epoch 1/1 -  Batch 167/1582 -  Discriminator loss: 0.3798 -  Generator loss: 4.3505 - 
Epoch 1/1 -  Batch 168/1582 -  Discriminator loss: 0.3795 -  Generator loss: 3.3628 - 
Epoch 1/1 -  Batch 169/1582 -  Discriminator loss: 0.3747 -  Generator loss: 4.1602 - 
Epoch 1/1 -  Batch 170/1582 -  Discriminator loss: 0.3699 -  Generator loss: 3.5044 - 
Epoch 1/1 -  Batch 171/1582 -  Discriminator loss: 0.3703 -  Generator loss: 3.5109 - 
Epoch 1/1 -  Batch 172/1582 -  Discriminator loss: 0.3732 -  Generator loss: 3.6101 - 
Epoch 1/1 -  Batch 173/1582 -  Discriminator loss: 0.3715 -  Generator loss: 3.8104 - 
Epoch 1/1 -  Batch 174/1582 -  Discriminator loss: 0.3733 -  Generator loss: 3.5085 - 
Epoch 1/1 -  Batch 175/1582 -  Discriminator loss: 0.3722 -  Generator loss: 3.7101 - 
Epoch 1/1 -  Batch 176/1582 -  Discriminator loss: 0.3706 -  Generator loss: 3.6365 - 
Epoch 1/1 -  Batch 177/1582 -  Discriminator loss: 0.3688 -  Generator loss: 3.5630 - 
Epoch 1/1 -  Batch 178/1582 -  Discriminator loss: 0.3763 -  Generator loss: 3.3629 - 
Epoch 1/1 -  Batch 179/1582 -  Discriminator loss: 0.3752 -  Generator loss: 3.7348 - 
Epoch 1/1 -  Batch 180/1582 -  Discriminator loss: 0.3756 -  Generator loss: 3.4044 -

Epoch 1/1 -  Batch 181/1582 -  Discriminator loss: 0.3763 -  Generator loss: 3.5319 - 
Epoch 1/1 -  Batch 182/1582 -  Discriminator loss: 0.3774 -  Generator loss: 3.8129 - 
Epoch 1/1 -  Batch 183/1582 -  Discriminator loss: 0.3732 -  Generator loss: 3.5189 - 
Epoch 1/1 -  Batch 184/1582 -  Discriminator loss: 0.3691 -  Generator loss: 3.6951 - 
Epoch 1/1 -  Batch 185/1582 -  Discriminator loss: 0.3665 -  Generator loss: 3.7146 - 
Epoch 1/1 -  Batch 186/1582 -  Discriminator loss: 0.3681 -  Generator loss: 3.8494 - 
Epoch 1/1 -  Batch 187/1582 -  Discriminator loss: 0.3687 -  Generator loss: 3.4952 - 
Epoch 1/1 -  Batch 188/1582 -  Discriminator loss: 0.3695 -  Generator loss: 3.8315 - 
Epoch 1/1 -  Batch 189/1582 -  Discriminator loss: 0.3692 -  Generator loss: 3.5292 - 
Epoch 1/1 -  Batch 190/1582 -  Discriminator loss: 0.3703 -  Generator loss: 3.5772 - 
Epoch 1/1 -  Batch 191/1582 -  Discriminator loss: 0.3730 -  Generator loss: 3.4389 - 
Epoch 1/1 -  Batch 192/1582 -  Discriminator loss: 0.3753 -  Generator loss: 3.8701 - 
Epoch 1/1 -  Batch 193/1582 -  Discriminator loss: 0.3772 -  Generator loss: 3.3972 - 
Epoch 1/1 -  Batch 194/1582 -  Discriminator loss: 0.3758 -  Generator loss: 3.6019 - 
Epoch 1/1 -  Batch 195/1582 -  Discriminator loss: 0.3764 -  Generator loss: 3.8490 - 
Epoch 1/1 -  Batch 196/1582 -  Discriminator loss: 0.3858 -  Generator loss: 3.0871 - 
Epoch 1/1 -  Batch 197/1582 -  Discriminator loss: 0.4024 -  Generator loss: 4.4664 - 
Epoch 1/1 -  Batch 198/1582 -  Discriminator loss: 0.3822 -  Generator loss: 3.2489 - 
Epoch 1/1 -  Batch 199/1582 -  Discriminator loss: 0.3757 -  Generator loss: 3.3509 - 
Epoch 1/1 -  Batch 200/1582 -  Discriminator loss: 0.3889 -  Generator loss: 4.2574 -

Submitting This Project¶

When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.